An aggressive (and delinquent) personality can be shown to exist when observations are made across a large enough sampling of exemplars. Arguments against the existence of such a personality are based on a failure to use sufficiently aggregated measures. Numerous estimates are preferable in order to average out idiosyncratic variance thereby leaving a clearer view of underlying dimensions of behaviour. We illustrate the usefulness of this principle and suggest that consistent patterns of individual and group differences in aggression are to be found, are associated with delinquent behaviour, and are moderately heritable.